Epoch AI | MirrorCode Benchmark: What's the largest software project AI can complete on its own?

🌟 How extensive of a software project can an AI system complete independently? Join ⁨Tom Adamczewski⁩ and Epoch AI as they delve into MirrorCode, a cutting-edge benchmark co-developed with METR designed for long-horizon coding tasks.

In this event, discover how AI must reimplement real programs end-to-end without access to their source code, successfully matching behavior on held-out tests. Covering 25 diverse programs across domains like Unix utilities, bioinformatics, interpreters, cryptography, and compression, we’ll explore:

• What insights MirrorCode provides on current AI coding abilities 🤖
• How to assess weeks-long autonomous AI work 📊
• The methodological challenges involved in gauging advancements in real software projects 🚧

✨ Learn more here

📅 Schedule

• 7:00 PM - Doors open with light refreshments 🥤
• 7:30 PM - Talk begins 🎤

Check out our other Summer Season events here ➡️